That looks really good, the complexity of modern K8s is incredible and assumes a high level of expertise, tools like Helm don’t really make it any simpler to maintain. I like the direction towards self-hosted solutions as well, vendor lock is not only dangerous for any business but can also be quite expensive. The only thing I don't understand here is the fundamental difference between Kamal and Ansible. And I would appreciate a comparison and the motivation that drove you to build your own instrument from scratch.
hey DHH, great stuff as always, looking forward to using this tool! I do wonder how you deal with logs at 37 signals. You demonstrated a request-based log search. What if you needed to retain the logs for a period of time or if you needed a more sophisticated search (maybe provided by the 3rd party already), how would you do that? Is that beyond the scope of this tool? Thanks
I find this intresting and is always good to have alternatives, but I don't see the added value. You can do the same with ansible or saltstack, the major "plus" is that is written in Ruby and compared to Cheff is insanely simple, but also Ansible. Again, I'm not trying to ditch it, just want to know whats the principle behind it and what makes it different from the alternatives (and yes, I read the blog about it), if we compare Puppet, Ansible, Saltstack and Cheff (the major players here) there are big differences and can be a hot debate.
Great stuff. Is there any plans to bump up the raw database server support to a more managed level database server, like taking care of regular offsite snapshots and reporting on issues with that process. I find most shops are failing this part which is imo a good reason for cloud managed databases. At the same time a process for migrating, putting sites on maintenance mode, moving the data with the provider move etc?
With Kubernetes I could specify resource requests and limits, to run multiple containers per server, and to schedule containers to run where the resources are available. The physical servers are huge, lots of CPU cores and memory. Does Kamal run a single container of your app per server? What if the servers are of different size: is it possible to configure it to run the specific number of the processes per container, or number of containers per server? Kubernetes limits also allow me to "sweep the memory leaks under the rug" - the bloated container will simply be killed and new ones started automatically. Kamal just doesn't seem to be enough for any serious production application by itself. There have to be other tools for auto-scaling, re-balancing load, etc. Amirite?
As a I understand kamal is only a deployment tool and not a scheduler right? If it so. What is the benefit instaed of using ansible cheff pupppet or terraform? If kamal is a workload scheduler what is the benefit against k3s or nomad? And why da hell rubby? 😂
Awesome, thanks so much for this. Really appreciate the time, tools, and knowledge you and your team continue to give to the community!
That looks really good, the complexity of modern K8s is incredible and assumes a high level of expertise, tools like Helm don’t really make it any simpler to maintain. I like the direction towards self-hosted solutions as well, vendor lock is not only dangerous for any business but can also be quite expensive.
The only thing I don't understand here is the fundamental difference between Kamal and Ansible. And I would appreciate a comparison and the motivation that drove you to build your own instrument from scratch.
Yet another banger! Thank you very much for what you give us, guys! 🔥
This is amazing, thank you so much.
Fantastic, simple and working
hi 🖐 With Postgres would be the same?
This is awesome. THANK YOU THANK YOU THANK YOU
Pure genius!, although I liked the name MRSK better.
Turns out the container logistics company with a similar name ALSO likes it.
MAERSK threatened to sue DHH for 13.7 billion dollars so he had to rename the tool to a random arab person's name instead.
Really?
I used Kamal, it works like a Swiss clock.
Very interesting and helpful. Thanks for Kamal and this video! 🥳😃👏
Great thank You!
hey DHH, great stuff as always, looking forward to using this tool! I do wonder how you deal with logs at 37 signals. You demonstrated a request-based log search. What if you needed to retain the logs for a period of time or if you needed a more sophisticated search (maybe provided by the 3rd party already), how would you do that? Is that beyond the scope of this tool? Thanks
They use Grafana + Loki + Promtail, K8s, Cloudwatch, S3, and DynamoDB for analysis.
@@GregRippetoe I thought they migrated off the cloud and don't use kubernetes. Do you have a source for that? thanks
@@emiribrahimbegovic813 They still use the cloud for some stuff but they might migrate fully after a while
It's crazy!!!!! I love it
did it contain tutorial for server without managed database?
I find this intresting and is always good to have alternatives, but I don't see the added value. You can do the same with ansible or saltstack, the major "plus" is that is written in Ruby and compared to Cheff is insanely simple, but also Ansible.
Again, I'm not trying to ditch it, just want to know whats the principle behind it and what makes it different from the alternatives (and yes, I read the blog about it), if we compare Puppet, Ansible, Saltstack and Cheff (the major players here) there are big differences and can be a hot debate.
Great stuff. Is there any plans to bump up the raw database server support to a more managed level database server, like taking care of regular offsite snapshots and reporting on issues with that process. I find most shops are failing this part which is imo a good reason for cloud managed databases.
At the same time a process for migrating, putting sites on maintenance mode, moving the data with the provider move etc?
With Kubernetes I could specify resource requests and limits, to run multiple containers per server, and to schedule containers to run where the resources are available. The physical servers are huge, lots of CPU cores and memory. Does Kamal run a single container of your app per server? What if the servers are of different size: is it possible to configure it to run the specific number of the processes per container, or number of containers per server? Kubernetes limits also allow me to "sweep the memory leaks under the rug" - the bloated container will simply be killed and new ones started automatically. Kamal just doesn't seem to be enough for any serious production application by itself. There have to be other tools for auto-scaling, re-balancing load, etc. Amirite?
What was wrong with kubernetes exactly?
Well done
Anyone can tell me how to connect Kamal to AWS EC2 with SSH?
is it based on mrsk ?
rename!
Nowonder the PHD in Kubernetes has been so elusive to me
Yo, kamal is here..
Nice 👍
It’s easy to do deployment. The hardest part is day 2 and patching/upkeep. Compliance is the hard part
Epic shit, will switch from capistrano/mina :)
I quite like thr name...:)
We need another one at normal speed 😂
Please change Kamal logs to not be red. This is too scary!
As a I understand kamal is only a deployment tool and not a scheduler right?
If it so. What is the benefit instaed of using ansible cheff pupppet or terraform?
If kamal is a workload scheduler what is the benefit against k3s or nomad?
And why da hell rubby? 😂
heroku going broke soon 😂
My dude is too fast even at 0.5 speed
millennial pause, sorry dhh
Nice 👍